Ph.D. Thesis Defense Announcement 

Develop Effective Graphitic Carbon Nitride Materials with the Help of Machine Learning and Design of Experiments 

by 

Liqing Yan 

Advisor(s): 

Dr. Yongsheng Chen 

Committee Members: Dr. Xing Xie; Dr. Sotira Yiacoumi; Dr. John Crittenden; Dr. Zhaohui Tong 

Date & Time: Monday, July 25th, 2022 at 1:00 PM (EST) 

Location: https://tinyurl.com/Liqing-defense

Complete announcement, with abstract, is attached 

Graphitic carbon nitride (g-C3N4) has attracted significant attention in energy and environmental applications. It is imperative to develop synthesis strategies that overcome shortcomings of g-C3N4 materials such as fast recombination of photoinduced electrons and holes, limited solar light utilization, and low specific surface area to expand its applications in the real world. Typically, optimization for the synthesis process is achieved through a one-variable-at-a-time method and only the photocatalytic activity is considered. The objective of this work is to advance the synthesis process of g-C3N4 materials for enhanced performance in environmental applications through a combined machine learning modeling and experiments approach.
Most studies on g-C3N4 materials are dedicated to improving its photocatalytic performance; however, an investigation of its adsorption performance not only expands its application scope but also provides insight into its surface interactions with molecules. Herein, sodium doped g-C3N4 (Na(x)-CN) is synthesized via a one-step thermal polycondensation approach. The prepared Na(x)-g-C3N4 demonstrates enhanced photodegradation efficiency of contaminants as well as superior adsorption capacities. Our work introduces a facile route to expand the application scope of g-C3N4.
To gain more systematic insights into the elemental doping process of g-C3N4, we have developed machine learning (ML) models from a large amount of literature data to link experimental parameters with the photocatalytic performance of element-doped graphitic carbon nitride (D-g-C3N4). The trained ML models are effective in predicting the photoactivity of D-g-C3N4 using experimental inputs. The method described in the present study provides valuable insights for the design of synthesis strategies for g-C3N4 before conducting time-consuming and expensive experiments.
Inspired by the ML model results and recent studies, hydrothermal pretreatment is adopted to prepare g-C3N4 nanosheets. Response surface methodology is utilized to guide experiments, evaluate the influence of synthesis conditions, and optimize the synthesis process. The photocatalytic activity and the yield of g-C3N4 materials whose optimal regions are distant from each other are simultaneously optimized using the desirability function approach. Our work provides valuable insights into the synthesis g-C3N4 with both high yield and improved efficacy, which is of great practical significance.
In summary, this thesis has successfully demonstrated the preparation of high-performing g-C3N4 materials with the help of machine learning modeling and the design of experiments. The novel Na-doped g-C3N4 expands the application scope of g-C3N4 for water remediation beyond the commonly studied photodegradation and provides insights into surface interactions between g-C3N4 and organic molecules. Moreover, the ML model sheds light on the ambiguous relationship between synthesis conditions and performance of elemental doped g-C3N4 and guides the synthesis of enhanced photocatalytic g-C3N4 materials. Finally, the simultaneous optimization of the yield and photocatalytic activity of g-C3N4 provides a roadmap to finding the conditions that are conducive to conflicting properties.